Adopting Digital Twins: Tackling industry barriers

3 min read

Dr Richard Ahlfeld, CEO and founder of Monolith AI, breaks down the barriers to augmenting a digital twin with AI and how industries can overcome them.

Landing in the top five in Accenture’s technology trends for 2021, digital twin technology is revolutionising design and manufacturing processes all over the world. However, its implementation across different industries can be stifled by barriers for engineers across these industries to truly utilise the true potential of this technology.

IBM defines a digital twin as “a virtual representation of an object or system that spans its lifecycle, is updated from real-time data, and uses simulation, machine learning and reasoning to help decision-making.” By creating a virtual model that is an exact counterpart of a physical object - be it a car or a bridge - you can analyse and test different scenarios to understand not only how a product performs, but how it will perform in the future under different conditions. The continuous collection and processing of data provides an objective, data-driven design that can be used to accelerate digital transformation across a range of sectors, such as design, manufacturing and even aeronautical engineering.

Beyond autonomy: how AI is driving the automotive industry

Previously only available to businesses with vast R&D teams, the recent advancements in data integration and general democratisation of machine learning has meant that this technology can become more widely used. Over time, the market for this technology has grown by 58 per cent annually and is expected to reach $48bn by 2026, saving companies resources, money and time. Recently, digital twin technology was used by NASA and the US government as part of their missile interceptor test, with Siemens also using it to solve heat dissipation problems caused by its radioisotope power generator. It has even moved away from traditional engineering sectors to be used for more everyday products, with Absolute Vodka for example, using the technology to design new products and streamline their manufacturing processes, from the supply chain through production and, eventually, to recycling and disposal.

In spite of the possibilities it provides, there are still some challenges to digital twin technology being adopted en masse, notably the complex software landscape and the level of expertise required to manage efforts. Most of the current academic research is focused on creating real-time models. There is some great insight on Neural Networks and how they can help describe really complex systems, but these are cumbersome to train and can quickly become complex without any physical interpretability and background knowledge on how to use them. The same can be said on the insight surrounding the creation of faster surrogate models from traditional methods such as Finite Element (FEA) or Computational Fluid Dynamics (CFD), which are widely used yet are too slow to get a complete understanding of complex physics in real-time. However, there is little research and insight available currently on software.

As data science becomes a bigger part of engineering design, engineering needs people who have experience with cloud technology and AI, but also who have significant and practical engineering experience.

digital twin technology
(Image: AdobeStock)

To overcome this, businesses can utilise machine learning to plug the gap, and pool existing data with engineers' insight to create a knowledge bank that can train itself. Honeywell’s Digital Transformation Officer Vincent Blake describes it well when he says: “What usually makes the difference is AI and subject matter expertise. It’s a marriage of data and expertise right down to the granular level."

Today, AI applications can be seamlessly integrated into existing engineering workflows - reducing valuable time spent and number of repetitive test cycles that are part of an engineer’s work. The result: better products getting quicker to the market, while staying ahead of the competition.

By adopting software engineers do not need to become expert coders and can introduce complex algorithms without having to adjust their core team. For example, Jota Sports Endurance Racing team are not software experts, but have extensively used software to create a digital twin of a car, which covers car setup, vehicle dynamics, aerodynamics and tires. By training deep learning models on car performance data, they are able to focus on making faster, better design decisions and streamline how their car and simulation data is validated.

Utilising software empowers the engineer without replacing but rather augmenting them, democratising its capabilities and freeing up time to focus on the creative, analytical and strategic work that can help drive the business, and industry forward.

Dr Richard Ahlfeld, CEO and founder of Monolith AI